resample_spatial:
Data like Sentinel-1, Sentinel-3 and Sentinel-5p is natively provided in sensor geometry. The transition to a rectangular grid with regularly spaced coordinates requires more advanced methods.
For instance, in sentinel-3, one method is "binning":
https://github.com/senbox-org/snap-engine/blob/1500b8ad538b6f4c97a722d25087cdea85fed504/snap-binning/src/main/java/org/esa/snap/binning/operator/BinningOp.java
Another method in sentinel-1 is referred to as "geocoding".
The most naïve approach here is to define new processes, but here I'm coining the idea of adding a custom resample method to the existing resample_spatial method.
Proposed solution:
Extend the list of resample_methods in resample_spatial.
If this idea is considered valid, I expect to be able to better define correct names and descriptions in the coming months. ('binning' and 'geocoding' or relatively generic terms, I'm still figuring out if we can be more specific.)
The alternative of defining new processes is possible, but I feel like resample_spatial is a more clear expression of what is happening, also allowing to capitalize on existing handling of this process in clients and backends. (It has a fairly specific effect on the spatial dimensions of a datacube.)
resample_spatial:
Data like Sentinel-1, Sentinel-3 and Sentinel-5p is natively provided in sensor geometry. The transition to a rectangular grid with regularly spaced coordinates requires more advanced methods.
For instance, in sentinel-3, one method is "binning":
https://github.com/senbox-org/snap-engine/blob/1500b8ad538b6f4c97a722d25087cdea85fed504/snap-binning/src/main/java/org/esa/snap/binning/operator/BinningOp.java
Another method in sentinel-1 is referred to as "geocoding".
The most naïve approach here is to define new processes, but here I'm coining the idea of adding a custom resample method to the existing resample_spatial method.
Proposed solution:
Extend the list of resample_methods in resample_spatial.
If this idea is considered valid, I expect to be able to better define correct names and descriptions in the coming months. ('binning' and 'geocoding' or relatively generic terms, I'm still figuring out if we can be more specific.)
The alternative of defining new processes is possible, but I feel like
resample_spatialis a more clear expression of what is happening, also allowing to capitalize on existing handling of this process in clients and backends. (It has a fairly specific effect on the spatial dimensions of a datacube.)